COMPMID-1727 - CL: Implement Gather

Change-Id: I3d859da09a4de1019bb8c2046725eab942247927
Reviewed-on: https://review.mlplatform.org/386
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>

3 files changed, 95 insertions, 2 deletions

diff --git a/arm_compute/core/CL/CLKernels.h b/arm_compute/core/CL/CLKernels.h
index cfcfb7400b..e68769b6ae 100644
--- a/arm_compute/core/CL/CLKernels.h
+++ b/arm_compute/core/CL/CLKernels.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2018 ARM Limited.
+ * Copyright (c) 2016-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -83,6 +83,7 @@
 #include "arm_compute/core/CL/kernels/CLGEMMReshapeLHSMatrixKernel.h"
 #include "arm_compute/core/CL/kernels/CLGEMMReshapeRHSMatrixKernel.h"
 #include "arm_compute/core/CL/kernels/CLGEMMTranspose1xWKernel.h"
+#include "arm_compute/core/CL/kernels/CLGatherKernel.h"
 #include "arm_compute/core/CL/kernels/CLGaussian3x3Kernel.h"
 #include "arm_compute/core/CL/kernels/CLGaussian5x5Kernel.h"
 #include "arm_compute/core/CL/kernels/CLGaussianPyramidKernel.h"
diff --git a/arm_compute/core/CL/kernels/CLGatherKernel.h b/arm_compute/core/CL/kernels/CLGatherKernel.h
new file mode 100644
index 0000000000..4dac6b0d1f
--- /dev/null
+++ b/arm_compute/core/CL/kernels/CLGatherKernel.h
@@ -0,0 +1,80 @@
+/*
+ * Copyright (c) 2018-2019 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef __ARM_COMPUTE_CLGATHERKERNEL_H__
+#define __ARM_COMPUTE_CLGATHERKERNEL_H__
+
+#include "arm_compute/core/CL/ICLKernel.h"
+#include "arm_compute/core/Types.h"
+
+namespace arm_compute
+{
+class ICLTensor;
+
+/** Interface for the kernel to perform tensor reshaping */
+class CLGatherKernel : public ICLKernel
+{
+public:
+    /** Default constructor */
+    CLGatherKernel();
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLGatherKernel(const CLGatherKernel &) = delete;
+    /** Prevent instances of this class from being copied (As this class contains pointers) */
+    CLGatherKernel &operator=(const CLGatherKernel &) = delete;
+    /** Allow instances of this class to be moved */
+    CLGatherKernel(CLGatherKernel &&) = default;
+    /** Allow instances of this class to be moved */
+    CLGatherKernel &operator=(CLGatherKernel &&) = default;
+    /** Default destructor */
+    ~CLGatherKernel() = default;
+    /** Initialise the kernel's inputs and outputs
+     *
+     * @param[in]  input   Source tensor. Supported tensor rank: up to 4. Data type supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
+     * @param[in]  indices Indices tensor. Supported tensor rank: up to 1. Must be one of the following type: S64. Each value Must be in range [0, input.shape[@p axis])
+     * @param[out] output  Destination tensor. Data type supported: Same as @p input
+     * @param[in]  axis    (Optional) The axis in @p input to gather @p indices from. Negative values wrap around. Defaults to 0
+     */
+    void configure(const ICLTensor *input, const ICLTensor *indices, ICLTensor *output, int axis = 0);
+
+    /** Static function to check if given info will lead to a valid configuration of @ref CLGatherKernel
+     *
+     * @param[in] input   Source tensor info. Supported tensor rank: up to 4. Data type supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32
+     * @param[in] indices Indices tensor info. Supported tensor rank: up to 4. Must be one of the following type: S64. Each value Must be in range [0, input.shape[@p axis])
+     * @param[in] output  Destination tensor info. Data type supported: Same as @p input
+     * @param[in] axis    (Optional) The axis in @p input to gather @p indices from. Negative values wrap around. Defaults to 0
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *indices, const ITensorInfo *output, int axis = 0);
+
+    // Inherited methods overridden:
+    void run(const Window &window, cl::CommandQueue &queue) override;
+
+private:
+    const ICLTensor *_input;   /**< Source tensor */
+    const ICLTensor *_indices; /**< Indices tensor */
+    ICLTensor       *_output;  /**< Destination tensor */
+    int              _axis;    /**< Axis index */
+};
+} // namespace arm_compute
+#endif /*__ARM_COMPUTE_CLGATHERKERNEL_H__ */
diff --git a/arm_compute/core/utils/misc/ShapeCalculator.h b/arm_compute/core/utils/misc/ShapeCalculator.h
index 4756ff4f97..ba0d8e254d 100644
--- a/arm_compute/core/utils/misc/ShapeCalculator.h
+++ b/arm_compute/core/utils/misc/ShapeCalculator.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2018 ARM Limited.
+ * Copyright (c) 2017-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -874,6 +874,18 @@ inline TensorShape compute_stack_shape(const ITensorInfo &a, unsigned int axis,
     }
     return shape_out;
 }
+
+inline TensorShape compute_gather_shape(const TensorShape &input_shape, const TensorShape &indices_shape, uint32_t actual_axis)
+{
+    ARM_COMPUTE_ERROR_ON(indices_shape.num_dimensions() > 1);
+    ARM_COMPUTE_ERROR_ON(input_shape.num_dimensions() > 4);
+    ARM_COMPUTE_ERROR_ON(actual_axis >= input_shape.num_dimensions());
+
+    TensorShape output_shape  = input_shape;
+    output_shape[actual_axis] = indices_shape[0];
+
+    return output_shape;
+}
 } // namespace shape_calculator
 } // namespace misc
 } // namespace arm_compute